globus-labs / mof-generation-at-scale / 8572042898

48%

push

github

Training fixes from test of workflow on Polaris (#98)

* Vary training set size

* Added ability to separate simulation and AI workers

* Add more options for parallel workflow

Fixes #77

* Compute the hosts afterwards

I need to learn class inheritance and dataclasses
or just go back to pydantic

* Bug fix: Ensure proper number of workers are started

* Override the environment

* Increase the range of the number of atoms in the prompt

* Remove deprecated field from test data

* Run at most N-1 nodes for simulation

* Make MOF queue LIFO not FIFO

* Turn off test epochs, update constants for MOFA loader

5 of 5 new or added lines in 3 files covered. (100.0%)

4297 of 10137 relevant lines covered (42.39%)

0.42 hits per line